1. Visualizing Non-Speech Sounds for the Deaf Tara Matthews Janette Fong, Jennifer Mankoff U.C. Berkeley Carnegie Mellon

2. UC Berkeley / Carnegie Mellon2 Motivation Ambient sounds give people an understanding of sounds relevant to their current situation or location: –serendipitous events (children playing in another room) –problematic things (faucet dripping) –critical information (fire alarm, knocking on the door) Maintaining this awareness can be difficult for people who are deaf

3. UC Berkeley / Carnegie Mellon3 Overview Study of peripheral, visual displays to help people who are deaf maintain an awareness of non-speech sounds Single IconSpectrograph with Icon

4. UC Berkeley / Carnegie Mellon4 Overview Study of peripheral, visual displays to help people who are deaf maintain an awareness of non-speech sounds Single IconSpectrograph with Icon

5. UC Berkeley / Carnegie Mellon5 Overview Study of peripheral, visual displays to help people who are deaf maintain an awareness of non-speech sounds Single IconSpectrograph with Icon

6. 6 Background Ho-Ching, Landay, Mankoff [CHI ‘03] –Gathered initial knowledge on visual displays of ambient sounds –Did almost no visual design exploration with users before implementing prototypes –Focused on quantitative comparison of two displays Spectrograph Positional Ripples (Map)

7. 7 Background Ho-Ching et al. [CHI ‘03] –Gathered initial knowledge on visual displays of ambient sounds –Did almost no visual design exploration with users before implementing prototypes –Focused on quantitative comparison of two displays Spectrograph Positional Ripples (Map)

8. 8 Background Ho-Ching et al. [CHI ‘03] –Gathered initial knowledge on visual displays of ambient sounds –Did almost no visual design exploration with users before implementing prototypes –Focused on quantitative comparison of two displays Spectrograph Positional Ripples (Map)

9. UC Berkeley / Carnegie Mellon9 Study Outline Design interviews (8 participants who are deaf) –Formal interview Understanding of user needs (sounds of interest, places of use, display size) –Design sketch interview Feedback on 10 sketches Visual design preferences & functional requirements

10. UC Berkeley / Carnegie Mellon10 Study Outline Implemented 2 fully-functional prototypes –Embody preferences & requirements found in interviews

11. UC Berkeley / Carnegie Mellon11 Study Outline Implemented 2 fully-functional prototypes –Embody preferences & requirements found in interviews Prototype evaluation (4 participants who are deaf) –In-lab prototype usage –Interview for feedback on prototypes

12. UC Berkeley / Carnegie Mellon12 Formal Interview Topics For people who are deaf, what are their preferred… Sounds to know about? –E.g., presence of others, phones, etc. Place for more sound awareness? –E.g., home, work, while mobile Display size? –E.g., PDA, PC monitor, large wall screen …in a visualization of non-speech sounds

13. UC Berkeley / Carnegie Mellon13 Formal Interview Results Users discussed a variety of sounds they wanted to know about…

14. UC Berkeley / Carnegie Mellon14 Results: Sounds of Interest Home –emergency alarms –wake-up alarms –doorbell and knocking –phone ringing –people shouting –intruders –children –items falling over –appliances (faucets dripping, water boiling, garbage disposal, gas hissing, etc.) “I need a [wake-up] alarm. Before an early flight, I will stay up all night.” “Once I left the vacuum cleaner on all night.” “I have trouble hearing my husband calling me… when I'm in another room”

15. UC Berkeley / Carnegie Mellon15 Results: Sounds of Interest Work –presence & activities of coworkers –emergency alarms –phone ringing –coworkers trying to get their attention –faxes/printers “My office is not ADA Compliant and I worry about missing the fire alarm.”

16. UC Berkeley / Carnegie Mellon16 Results: Sounds of Interest Walking / running outside –dogs barking –honking –vehicles –bikes or people coming up behind them –if blocking another person In vehicle –cars honking –sirens –sounds indicating problems with the car “When I first moved to L.A. I was surprised at how some drivers are aggressive on the roads and at intersections. I had some close calls.” “ When there is something wrong with the car… it tends to go unnoticed until it is very expensive to fix. ”

17. UC Berkeley / Carnegie Mellon17 Formal Interview Results Places for more sound awareness –Wanted to be more aware of sounds in all places –Especially at home, work, in the car, & while walking Display size –Preferred small displays (PDA or part of a PC screen) –Large wall screens also valued at home (better visibility)

18. UC Berkeley / Carnegie Mellon18 Design Sketch Interview Topics For people who are deaf, what are their preferred… Information about sounds? –E.g., sound recognition, location, characteristics (volume & pitch) Visual design characteristics? –E.g., shapes & colors, pictorial, text-based, graphs Functions? –E.g., customization, etc. …in a visualization of non-speech sounds

19. UC Berkeley / Carnegie Mellon19 Design Sketches Designs varied in visual design and information conveyed recognized sounds, location, volume & pitch LED Panels Directional IconsMap

20. UC Berkeley / Carnegie Mellon20 Design Sketches Designs varied in visual design and information conveyed recognized sounds, location, volume & pitch LED Panels Directional IconsMap

21. UC Berkeley / Carnegie Mellon21 Design Sketches Designs varied in visual design and information conveyed recognized sounds, location, volume & pitch LED Panels Directional IconsMap

22. UC Berkeley / Carnegie Mellon22 Design Sketches Designs varied in visual design and information conveyed recognized sounds, location, volume & pitch LED Panels Directional IconsMap

23. UC Berkeley / Carnegie Mellon23 Ambient Spectrograph with Recognition Bubbles Sidebar “I think using the Spectrograph will be useful, so I can glance at it and figure out the sound.” It “looks like a piece of art on the walls… I wouldn’t mind having it hanging in the living room.” “I’d have to practice and learn this to understand it.” “It doesn't identify WHAT noise is being made and if it's an important noise… I would prefer some sort of identification of the noise that is being made.”   Results

24. UC Berkeley / Carnegie Mellon24 Ambient Spectrograph with Recognition Bubbles Sidebar “I think using the Spectrograph will be useful, so I can glance at it and figure out the sound.” It “looks like a piece of art on the walls… I wouldn’t mind having it hanging in the living room.” “I’d have to practice and learn this to understand it.” “It doesn't identify WHAT noise is being made and if it's an important noise… I would prefer some sort of identification of the noise that is being made.”   Results

25. UC Berkeley / Carnegie Mellon25 Ambient Spectrograph with Recognition Bubbles Sidebar “I think using the Spectrograph will be useful, so I can glance at it and figure out the sound.” It “looks like a piece of art on the walls… I wouldn’t mind having it hanging in the living room.” “I’d have to practice and learn this to understand it.” “It doesn't identify WHAT noise is being made and if it's an important noise… I would prefer some sort of identification of the noise that is being made.”   Results

26. UC Berkeley / Carnegie Mellon26 Ambient Spectrograph with Recognition Bubbles Sidebar “I think using the Spectrograph will be useful, so I can glance at it and figure out the sound.” It “looks like a piece of art on the walls… I wouldn’t mind having it hanging in the living room.” “I’d have to practice and learn this to understand it.” “It doesn't identify WHAT noise is being made and if it's an important noise… I would prefer some sort of identification of the noise that is being made.”   Results

27. UC Berkeley / Carnegie Mellon27 Design Sketch Interview Results Information about sounds –sound recognition Visual design characteristics –easy to interpret –glanceable –appropriately distracting Functions –identify what sound occurred –view a history of displayed sounds –customize the information that is shown –determine the accuracy of displayed information

28. UC Berkeley / Carnegie Mellon28 Using Interview Results: Prototypes Prototyped two designs –Single Icon, Spectrograph with Icon Incorporated support for desired functions –sound identity, customization, accuracy, history Sound recognition –Used existing, state-of-the-art recognition system [Robert Malkin, HSCMA ’05] –Used audio only to detect and classify events based on training

29. UC Berkeley / Carnegie Mellon29 Single Icon Identity: icon shows recognized sound; rings for unrecognized Customization: select sounds to show in menu Accuracy: icon opacity and text indicate sureness of recognition system History: graph of past sounds Size: small (55 x 93 pixels)

30. UC Berkeley / Carnegie Mellon30 Single Icon Identity: icon shows recognized sound; rings for unrecognized Customization: select sounds to show in menu Accuracy: icon opacity and text indicate sureness of recognition system History: graph of past sounds Size: small (55 x 93 pixels)

31. UC Berkeley / Carnegie Mellon31 Single Icon Identity: icon shows recognized sound; rings for unrecognized Customization: select sounds to show in menu Accuracy: icon opacity and text indicate sureness of recognition system History: graph of past sounds Size: small (55 x 93 pixels)

32. UC Berkeley / Carnegie Mellon32 Single Icon Identity: icon shows recognized sound; rings for unrecognized Customization: select sounds to show in menu Accuracy: icon opacity and text indicate sureness of recognition system History: graph of past sounds Size: small (55 x 93 pixels)

33. UC Berkeley / Carnegie Mellon33 History Display Colored bars represent recognized sounds –Sound  color –Volume  bar height –Time  x-axis position

34. UC Berkeley / Carnegie Mellon34 Spectrograph with Icon Combines Single Icon with a Spectrograph Identity, customization, accuracy, history: same as Single Icon Adds ability for in-depth interpretation of sounds: –amplitude  darkness –frequency  y-axis –time  x-axis Size: small-medium (263 x 155 pixels)

35. UC Berkeley / Carnegie Mellon35 Prototype Evaluation Evaluated with 4 users who are deaf Recognition system set-up –Trained with office sounds phone ringing, voices, door opening/closing and knocking. –Filtered out background noises typing, mouse clicks, chair creaks, and continuous background noises (e.g., heaters & fans). –One high-quality microphone (Sony ECM 719), mounted on wall, above & behind the desk/PC

36. UC Berkeley / Carnegie Mellon36 Prototype Evaluation In-lab prototype usage –User checked email & monitored display on PC –Researcher created sounds: phone ringing, door opening/closing, coughing, pen clicking, shuffling papers, knocking on the door, voices, banging a plastic toy against the table, and hitting a metal bookcase Interview for feedback on prototypes

37. UC Berkeley / Carnegie Mellon37 Prototype Evaluation Results User preferences, in order: 1.History Display (as a stand-alone interface) 2.Single Icon 3.Spectrograph with Icon

38. UC Berkeley / Carnegie Mellon38 Results: History Display Enabled users to know what sounds occurred w/o constantly watching display Alerted users to interesting sounds –louder sounds made bigger bars, attracting attention

39. UC Berkeley / Carnegie Mellon39 Results: Single Icon Overall, positive reactions: –enabled users to recognize sounds Suggested improvements: –better visual distinction between important and unimportant sounds –rings not sufficient for unrecognized sounds: needed more information (e.g., location)

40. UC Berkeley / Carnegie Mellon40 Results: Spectrograph with Icon Mixed reactions: –confusion difficult to interpret spectrograph hard to distinguish important sounds –interest exploring shapes caused by sounds Suggested improvements: –display icon or text over the spectrograph

41. UC Berkeley / Carnegie Mellon41 Summary Study of peripheral, visual displays to help people who are deaf maintain an awareness of non-speech sounds Contributions: 1.Understanding of user needs sounds of interest, places of use, display size Visual design preferences easy to interpret, glanceable, appropriately distracting Functional requirements sound identity, customization, accuracy, history 2.Two fully functioning prototypes designed, implemented, and evaluated embody preferences and requirements in (2)

42. UC Berkeley / Carnegie Mellon42 Questions? For more information: tmatthew@cs.berkeley.edu www.eecs.berkeley.edu/~tmatthew/projects/ic2hear.html Thank you! This work was supported by NSF grants IIS-0209213 and IIS-0205644

43. UC Berkeley / Carnegie Mellon43 Recognition System Accuracy Phone100% Voices100% Door knock100% Door open / close76%